DeepSeek is about to release a new model and is still deeply adapting its underlying computing power with Huawei, which has delayed a lot of progress. It is said that inference and training will all run on Huawei chips. If this matter comes to fruition, the script that Huang Renxun was most worried about will come true. He mentioned in an interview with Dwarkesh Patel a few days ago: 'If one day DeepSeek level models are first released on Huawei chips, it will be a terrible result for our country.'. '' This' someday 'may be coming soon. The logic behind Huang Renxun's opposition to chip export controls is simple (of course, I think he still wants to make money from China, which is the largest market outside of the United States) 1. China's computing power is already sufficient. AI training is a parallel computing problem, and what an H100 can do, a stack of 7nm chips can also do. China has a large amount of 7nm production capacity and cheap energy, and Anthropic's Mythos was trained on a "fairly ordinary computing power scale". This computing power already exists in China, indicating that the top models have been trained. China's computing power is already sufficient. Huawei's revenue in 2025 is 880.9 billion yuan, and millions of chips have been shipped. 2. Algorithms determine the upper limit more than computing power. China has over 50% of the world's AI researchers, just take a look at the world's top AI laboratories, where the majority are of Chinese descent. DeepSeek was not built by stacking cards, it is a breakthrough at the algorithmic level. Computing power is the lower limit, algorithms are the upper limit. The training cost of Chinese models is less than one tenth of that of the United States, and the electricity cost of Chinese computing centers is only half of that of the United States. Therefore, the current pricing of Chinese model APIs is not selling at a loss, and the profit margin may not be as large as we imagine compared to the United States. If the model capability is the same as that of the United States, but the cost is only one-fifth, that would be the biggest nightmare. Just like the quality of goods made in China is the same as in the United States, but the cost is less than one-third of that in the United States. The answer to whether users will vote in the end is very clear. 3. The real effect of the ban is to help Huawei maintain its ecosystem. Unable to buy Nvidia → Can only use domestic chips → Software ecosystem revolves around domestic chip adaptation → Huawei chips are becoming more and more user-friendly → Even if the ban is lifted, customers may not necessarily come back. China accounts for 40% of the global technology industry, and voluntarily abandoning this market is not protection, but self harm. Huang Renxun's original words: "If AI models run best on other people's technology stacks in the future, that will be the nightmare for the United States. '' So what? The moat of the chip industry is not only technological leadership, but also ecological lock-in. Customers write code around your architecture, and your moat will become deeper and deeper. Customers will be forced to turn to other architectures, and your moat will be filled. China spent 3 years migrating, and it is foreseeable that it will be implemented in the near future. If DeepSeek really completes the full chain verification from training to inference on Huawei chips, it means that the Chinese AI ecosystem has run a completely independent technology stack from chips to frameworks to models for the first time. The pitfalls that DeepSeek and Huawei chips have overcome can be reused on other Chinese open-source models, as well as domestically produced chips. All corresponding ecosystems will keep up, and it may not take three years to catch up with the current ecosystem in the United States. I think this paragraph is quite reasonable. The development of the Internet in China is comparable to that in the United States because of the existence of walls. If there is no wall, China should have all the Internet products of the United States by now. Maybe the domestic Internet will not have so much space. China today is not catching up, it is diverging. What Huang Renxun saw was not that China was chasing, but that China was starting a new stove. And the ban is the driving force that forces them onto this path.